By Andy Pease, President and CEO of QuickLogic
The Early Years of the Semiconductor Industry
When the semiconductor industry was born approximately fifty years ago, the biggest problem it faced was how to reliably and economically manufacture large numbers of silicon-based devices. Because this was a new industry, there was no eco-system. This forced early device manufacturers to develop their own equipment to support the entire manufacturing chain from wafer fabrication to assembly to test. These companies had to be vertically integrated, as they had no other choice.
The limited capabilities of the early technology meant that manufacturers only could create very simple circuits. For example, at that time, state of the art for analog circuits was an op amp with 20 transistors. Digital logic was also in its infancy, with a quad two-input NAND gate the most leading edge device then available. In retrospect, this was definitely “Small Scale Integration”.
One of the consequences of the limited capability provided by the early devices was that system design, and even sub-system design, was carried out by OEMS who directly bought and used the devices. They also were forced to develop their own tools and processes since there was no system design ecosystem.
About thirty-five years ago, the picture began to change. Large Scale Integration (LSI) came onto the scene and the capabilities of semiconductor manufacturing had improved to the point where small systems, such as the 4004, could be implemented on a single piece of silicon. Now that system design was directly applied to the silicon devices, the US government started to push for Very Large Scale Integration (VLSI). Carver Mead recognized that semiconductor complexity had matured to the point that design was the gating element. This started a revolution which blurred the line between designers and manufacturers.
Around the same time and perhaps even a bit earlier, elements of semiconductor manufacturing slowly begun to be outsourced. The initial efforts were focused on assembly as new equipment became available. This enabled independent operators using less expensive labor to provide that service more economically than the vertical manufacturers. Testing the devices coming out of assembly in a production environment was the next logical step to be outsourced for many of the same reasons.
Wafer fabrication, the most complex and capital-intensive step in the semiconductor manufacturing business, was the last bastion to fall. However, as semiconductor companies continued the relentless drive for ever smaller process geometries, the complexity and cost of wafer manufacturing continued to climb until only the largest companies could afford to keep up with the state of the art in the industry. Those of us who have been around long enough remember the debate between the large company executives who believed that “real men have fabs” and the response from the smaller company executives that “no – real men make money”.
The Fabless Model and Its Implications
Fortunately for the smaller companies, around 25 years ago, third party wafer fabricators (“fabs”), led by TSMC, began offering wafer manufacturing services. These companies focused their attention on raising the enormous amounts of capital required to establish a fab and operate it efficiently. They could also afford to hire many of the best and brightest engineers and thus invest the human capital necessary to stay on the leading edge of technology.
Suddenly it became possible to be a semiconductor company without having to own and operate a fab. New “fabless” semiconductor companies sprang up to exploit this capability, which freed them to focus their time and energy on providing more complete solutions to their end customers.
Initially, “more complete solutions” meant silicon devices with ever-increasing levels of complexity and capability. Analog and digital semiconductors slowly evolved from small design elements to sub-systems and ultimately to complete “systems on a chip” (SOCs).
While this was happening, companies began to extend their product solutions beyond the devices themselves. For example, programmable logic companies such as QuickLogic realized that their customers would require design implementation software and a well-trained team of applications engineers to support them. Over time this software became quite sophisticated in its own right, and an inseparable part of the product solution offered by the company.
This trend was reflected across the industry. Former “semiconductor manufacturing” companies began to provide a much broader horizontal product solution which often included not just silicon, or even silicon plus software, but development boards and kits, algorithms, integrated off-the-shelf or customer-specific IP, and a host of support services.
The Fabless Semiconductor Association is Born
The fabless semiconductor revolution helped drive increasing complexity in the semiconductor devices themselves as well as their surrounding ecosystem. As more and more companies jumped on the fabless bandwagon, a clear need for a new industry association developed. As a result, about twenty years ago, the precursor to the GSA, known as the “Fabless Semiconductor Association”, or FSA, was born. A handful of companies such as QuickLogic became the founding members of the FSA.
At first, the value to each member was the ability to band with other fabless companies and to share information about the foundry service providers. As the fabless model gained acceptance, the FSA acted as the flag bearer by giving all of its members more leverage than any of them could obtain on their own. One focus of the association was to help balance wafer demand with wafer capacity, an early challenge of the fledgling fabless industry.
During its early years, the association brought together suppliers, fabless companies, integrated device manufacturers (IDMs) and other organizations that play a vital role in the semiconductor industry. Shortly afterwards, some IDMs such as United Microelectronics Corporation (UMC), announced their intention to become pure-play silicon foundries, further establishing the strong trend towards the fabless model.
Since that time, the pace of change has continued relentlessly. As the table below (from IC Insights) shows, revenue from fabless semiconductor companies as a percentage of total industry revenue has steadily grown from single digits to almost 30% in 2013. That 30% translates into nearly $78 Billion in revenue. Equally significantly, TSMC, the leading pure-play silicon foundry has become a recognized leader in developing semiconductor technology. In fact, the fabless model has become so well established that in 2007 the association changed its name from the FSA to the “Global Semiconductor Alliance”, or “GSA”, to reflect the fact that its fabless mission had been achieved. Now it was time to take on a broader role as a global semiconductor trade organization.
The GSA’s current mission is not only to continue to foster the fabless ecosystem, but to help drive the growth and economic success of the entire global semiconductor industry supply chain. In line with the evolution of semiconductor companies from point silicon providers to complete product solution providers, the new organization includes wafer manufacturing, assembly, and test as well as critical elements of the support ecosystem, including electronic design automation (EDA) and design software tools and intellectual property (IP).
Alongside the industry and the GSA, QuickLogic has continued to evolve, from an FPGA company to a Customer Specific Standard Product (CSSP) company with a current focus on always-on, always-aware sensor hubs for smartphones, tablets and wearables. Each step in its evolution has meant a further increase in the scope of its product offering which now includes silicon, software, soft and hard IP, development boards and kits, algorithm libraries, technical support and verified third-party device interoperability.
As a survivor of twenty-five years of relentless change, QuickLogic is proud to continue its association with the GSA. While the core product produced by the members of the GSA will continue to be semiconductors, the likely future for all of us is one where system software and other elements of the complete product solution play an increasing role in product differentiation. No doubt these changes will continue with adaptable companies and organizations like QuickLogic and the GSA leading the charge.